Re: lexical resources with n-ary translations

Hi all,

Sorry, I was not available for the last telco.

As Christian said correctly, there are indeed many dictionaries with more than 2 languages. However, at least for traditional printed(/printable) dictionaries, they are “multi-target fork dictionaries" (as I called them in my thesis). They have one source language and several target languages. Sometimes, you have “multi-target linked directory” as the French -> English -> Malay” dictionary we produced decades ago.

As far as terminologies are concerned, you may find “tables” where there is no (official) source languages (one language by column, one concept by line), but this is only possible when you have not meaning shifts between your set of languages in the terminology. Not sure you can represent such terminologies using vartrans.

For more complex domains (as we did with legal domain for the Alpine convention during the LexALP project), we used the “axies” based lexical organisation proposed in my thesis and it was quite successful (and it solved some of the challenges terminologists were confronted to before).

Regards,

Gilles,

> On 26 Jun 2020, at 10:03, Christian Chiarcos <christian.chiarcos@web.de> wrote:
> 
> Dear all,
> 
> as an afterthought to the discussion on n-ary translations, I can see that MT translation tables for more than two languages are somewhat artificial. However, (print) dictionaries with more than one target translations are actually very common in the technical domain, see https://www.springer.com/de/book/9789020116670 <https://www.springer.com/de/book/9789020116670> as an example. For a possible revision of OntoLex core, it might thus be worth considering to drop the functional restriction for vartrans:target (but not for vartrans:source).
> 
> Best,
> Christian